To compare the diagnostic ability of somatosensory evoked potentials and the number connection test in the detection of subclinical hepatic encephalopathy, 44 decompensated cirrhotic patients with no overt encephalopathy were admitted to the study and followed up for 6 months. Twenty-two patients developed overt encephalopathy during follow-up, and were retrospectively considered as patients with subclinical encephalopathy at entry. Of the 15 patients with abnormal somatosensory evoked potentials, 14 (93%) experienced at least one episode of hepatic encephalopathy in the following 6 months, while 16 (70%) of the 23 patients with abnormal number connection test did so. The data suggested that somatosensory evoked potential measurement was less sensitive, but more specific than number connection test in the detection of subclinical encephalopathy, although the differences were not statistically significant. Besides, the number connection test is much simpler, convenient and cheap, and it specificity could be improved if the learning effect and age, education and visual factors were carefully considered. We therefore recommend using the number connection test routinely in patients with decompensated liver diseases for the detection of subclinical hepatic encephalopathy, and reserving somatosensory evoked potential measurement as a complementary examination, when appropriate, or for assessment of overt hepatic encephalopathy.